in the nimzo

why the variation 4.Bf4 is never played in the nimzo indian although it's logical and the computer likes it?

Because in a lot of pawn structures that could potentially come out of the Nimzo Indian, the bishop doesn't make too much sense on f4.
For example, if black plays c5, d5 & cxd4 and white eventually ends up with an isolated queen's pawn, the bishop would belong on g5 rather than f4.
Or if black goes d6 & e5, the bishop on f4 will have to move again.

So, white typically plays more flexible moves first, and only much later decides where he's going to put that queen's bishop. (This idea of waiting with committing a certain piece and playing more flexible developing moves first is actually quite common, especially with regard to the queen's bishop. You might have heard the saying "develop your knights before bishops", which is basically an oversimplification of this theme.)

If you want to bring out the bishop immediately, then g5 is probably the best square for it because by pinning the f6 knight, it indirectly helps with controlling the e4 square, which tends to be important in the Nimzo.

Also note that certain motifs that might sometimes occur in other 1. d4 openings with a bishop on f4, are less likely to happen in the Nimzo Indian: For example, Nc3-b5, threatening Nc7, will pretty much never be an idea in the Nimzo as the c3 knight either cannot move due to the pin, or will be taken out before Nb5 ever becomes a problem.

thank you very much@Dr_King_Schultz for the answer . But you have also take a look about the advantages of this move.I think in fact black can not achieve the plan d6 and e5 and it invites black to take the c3 knight then white have the two bishops. I checked the Bf4 move and all computers prefer this move to Bg5 !!

White gets the 2 bishops in most Nimzo Indian lines, so that's nothing special about 4. Bf4. Usually, black will get compensation for the bishop pair either in the form of better pawn structure or better development.
And the d6 & e5 plan should be quite easy to carry out if black wants as he has a lot of ways to support this: With Nc6, 0-0 & Re8 and Qe7, he can organize 3 pieces (plus the d6 pawn) to support the e5 break in a very natural way, whereas white only has the knight on f3 and the bishop on f4 (plus the pawn on d4) contesting the e5 square. And even if he manages to somehow get the queen to defend against the e5 break (for example, by playing Qc2 followed by a3 Bxc3 Qxc3 at some point), he still won't have enough control over e5 to stop this advance.
Btw, I'm not saying that 4. Bf4 is a terrible move; I'm just trying to explain why it's not that popular. ;)

The argument of "knights first" can also be made against the London: 1 d4 Nf6 2 Bf4. I guess that the main reason is that in the London system the pawn goes to c3, not c4. The London system is kind of a reversed Lasker defence to the Reti opening: 1 Nf3 d5 2 g3 Bf5 3 c4 c6 4 Bg2 h6.

It's not bad but no remedy against everything. For example it does nothing against the pinning and doubling on c3 and it doesn't pin the Nf6 (-e4) itself.

It's a just a developing move with a diffuse purpose.

I played 4 ... b6 (Bb7, Ne4, Bxc3) with relative good play as far I remember.

it doesn't pin the Nf6 (-e4) itself
That can also be said in the orthodox queen's gambit declined: 1 d4 d5 2 c4 e6 3 Nc3 Nf6 where 4 Bg5 was obligatory to pin the knight, but 4 Bf4 got popular too.